Predicting Session Attendance at the 2005 JSM

• Activity Number: 191
• Type: Contributed
• Date/Time: Monday, August 8, 2005 : 2:00 PM to 3:50 PM
• Sponsor: General Methodology
• Author(s): Ye Zhong*+ and Michael R. Elliott and John Kolassa and Daniel Heitjan
• Companies: University of Pennsylvania and University of Pennsylvania and Rutgers University and University of Pennsylvania
• Keywords: negative binomial regression ; predictive model ; attendance

Abstract

The staff of the Joint Statistical Meetings (JSM) attempt to assign sessions to meeting rooms they will fill comfortably but not overfill. The current system involves eliciting room attendance from the JSM Program Committee member responsible for the session and combining their predictions with traditional wisdom, but often there are errors. We are not aware of any past systematic attempts to predict room sizes. In an effort to develop an improved system for producing predictions, we analyzed data from the 2003 (San Francisco) and 2004 (Toronto) JSMs. Data included a number of descriptors of session type, time, sponsor with Program Committee member's projected attendance, and the actual attendance. We used negative binomial regression to develop a predictive model from 2003 JSM data. The time, type, sponsor, and the PC member's prediction were all significant factors. We are in the process of applying the model to predict 2004 JSM attendance data and compare the quality of these predictions to those from the 2004 Program Committee members. We will use a similar methodology to predict attendance at the 2005 JSM.